[DeepSpeed] add weight_mapping to _load_state_dict_into_zero3_model

[kashif]

What does this PR do?

This pull request introduces enhancements to the DeepSpeed integration for model loading, specifically improving how weight conversions (such as renaming and merging/splitting of weights) are handled when loading state dictionaries into models using DeepSpeed ZeRO-3. The changes add a new utility for applying weight conversion logic before loading, and update the model loading pipeline to use this utility when needed.

Key changes:

Weight Conversion Utilities

• Added a new function _apply_weight_conversions_to_state_dict in deepspeed.py to handle weight renaming and merging/splitting operations on state dicts before loading them into a model. This function supports both simple renaming and more complex conversions using WeightConverter and WeightRenaming objects.

DeepSpeed Model Loading Logic

• Updated _load_state_dict_into_zero3_model to accept an optional weight_mapping argument. If provided, it uses the new _apply_weight_conversions_to_state_dict function to preprocess the state dict before loading. It also stores the applied weight conversions on the model for later reference.
• Modified the call to _load_state_dict_into_zero3_model in _load_pretrained_model (in modeling_utils.py) to pass the weight_mapping argument, ensuring that weight conversions are applied during DeepSpeed ZeRO-3 model loading.

Fixes #43257

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you read the contributor guideline,
  Pull Request section?
• Was this discussed/approved via a Github issue or the forum? Please add a link
  to it if that's the case.
• Did you make sure to update the documentation with your changes? Here are the
  documentation guidelines, and
  here are tips on formatting docstrings.
• Did you write any new necessary tests?

Who can review?

Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[github-actions]

View the CircleCI Test Summary for this PR:

https://huggingface.co/spaces/transformers-community/circle-ci-viz?pr=43303&sha=8b03dd

[ArthurZucker]

This is great thanks for the initiative!

• will have device issues probably no?
• for experts, if deepspeed places some on device x,y,z merging or etc could break this
• memory consumption issue!

[ArthurZucker]

Just merged #43261 which changes stuff a but I'll check this PR out to see if we can simplify a bit calls to re-use some stuff on it today

[kashif]

i'll use the new LoadStateDictConfig

[ArthurZucker]

LGTM